reqs.md

Reqs for insitucor:

insitucor calls the complete workflow for spatial correlation analysis.

Inputs:

• counts Single cell expression matrix. Normalizing the data to give every cell the same total expression is preferred.
• conditionon Data frame of variables to be conditioned on when computing gene correlation. Rows correspond to the rows of counts and xy. At a minimum, it is recommended to include cell type and tissue ID. Including cell signal strength (total counts) and background (negmean) is also recommended.
• celltype Vector of cell type assignments
• neighbors Neighbor relationships, stored as a sparse matrix
• xy Matrix of xy coordinates.
• k k for k-nearest neighbor network building
• radius Radius for neighbor network building
• tissue Used for neighbor network building. Neighbors will only be considered for cell with the same tissue value.
• mincor Correlation values below this threshold will be stored as 0 to allow for a sparse matrix
• resolution Argument to igraph::cluster_leiden. Lower values produce bigger clusters.
• corthresh Only correlations about this value will go into the adjacency graph fed into leiden clustering
• min_module_cor Only keep modules with average cor above this value.
• max_cells If there are more than this many cells, certain steps will use a random subset of this size.

Outputs:

A list, with the following elements: \enumerate{ \item condcor: A sparse matrix holding genes' conditional correlations. \item modules: A data frame detailing gene module membership and weights \item scores_env: A matrix of cell * module environment scores \item scores_sc: A matrix of cell * module single cell scores \item attributionmats: A list of matrices holding attribution scores for each cell type * gene in each module. \item celltypeinvolvement: A matrix giving the maximum attribution score for each cell type in each module. }

Reqs for calcSpatialCor:

All the steps to compute the conditional spatial correlation matrix.

Inputs:

• counts Single cell expression matrix. Normalizing the data to give every cell the same total expression is preferred.
• conditionon Data frame of variables to be conditioned on when computing gene correlation. Rows correspond to the rows of counts and xy. At a minimum, it is recommended to include cell type and tissue ID. Including cell signal strength (total counts) and background (negmean) is also recommended.
• neighbors Neighbor relationships, stored as a sparse matrix
• xy Matrix of xy coordinates
• k k for k-nearest neighbor network building
• radius Radius for neighbor network building
• tissue Used for neighbor network building. Neighbors will only be considered for cell with the same tissue value.
• roundcortozero Correlation values with absolute values below this threshold will be stored as 0 to allow for a sparse matrix. If set to NULL, then no rounding to 0 will happen, and a dense matrix will be returned.
• max_cells If there are more than this many cells, certain steps will use a random subset of this size. Output will still be for all cells.
• verbose Whether to print progress

Outputs:

A list, with the following elements: \enumerate{ \item condcor: The conditional correlation matrix \item env: The matrix of neighborhood expression of all genes (columns) in all cells - or a subset of cells (rows) \item neighbors: a sparse adjacency matrix giving neighbor relationships }

Reqs for defineModules:

Define gene modules given the spatial correlation matrix.

Inputs:

• condcor Conditional correlation matrix
• env Environment matrix, possibly for a large subset of cells
• min_module_size Modules smaller than this are discarded
• max_module_size Modules bigger than than this are subclustered
• gene_weighting_rule How to define modules' gene weights, absed on gene expression levels. One of "inverse_sqrt", "inverse", or "identity".
• resolution Resolution parameter for leiden clustering
• corthresh Correlations with absolute value below this will be rounded to zero to save memory
• min_module_cor Modules must have mean correlation of at least this much to be reported

Outputs:

A list, with the following elements: \enumerate{ \item modules: A list giving the gene names in each module \item weights: A list giving the gene weights in each module \item weightsdf: A data frame summary of genes' module membership and weights }

Reqs for scoreModules:

Score gene modules over single cells and their environments.

Inputs:

• counts Single cell expression matrix
• weights List of module weights, as output by defineModules
• neighbors Adjacency matrix of neighbors

Outputs:

A list, with the following elements: \enumerate{ \item scores_env: A matrix of module scores for cell environments \item scores_sc: A matrix of module scores for single cell profiles }

Reqs for cellTypeAttribution:

Score cell types' involvement in modules

Inputs:

• modulescores Matrix of each cell's score for each module.
• weights List of module gene weights.
• counts Single cell expression matrix, cells in rows, genes in columns. The same data used in earlier steps.
• celltype Vector of cell types
• neighbors Neighbor relationships, stored as a sparse matrix
• nsub Subsample size to use. This function only extracts correlations, so 5000 cells is ample
• verbose Whether to print progress

Outputs:

A list, with the following elements: \enumerate{ \item "involvescores", a matrix of module * cell type scores, from 0-1. \item "attributionmats", a list of matrices giving the involvement of each cell type in each gene in each cluster. }

Reqs for plotCorrelationNetwork:

Make an igraph plot of the network structure of the conditional correlation matrix:

Inputs:

• x Correlation matrix
• modules Either the "modules" data frame output by sparc(), or a vector of module names.
• genes If modules is given as a vector, then specify the vector of genes corresponding to it.
• corthresh Connect genes with cor > this value
• show_gene_names Logical
• vertex_size Argument passed to igraph

Outputs:

Draws an igraph network plot